benchgap
Alibaba (Qwen) · model

Qwen3 235B 2507 benchmark scores

As of 2026-11-22, Qwen3 235B 2507 (Alibaba (Qwen)) has measured scores on 9 benchmarks and estimated scores on 37 more.

BenchmarkScoreSource
AA-GPQA Diamond83.1%estimated ± 3.7 pp, high confidence
AA-HLE13.7%estimated ± 5.7 pp, medium confidence
AA-SciCode17.5%estimated ± 4.7 pp, medium confidence
AIME2671.4%estimated ± 4.3 pp, medium confidence
Artificial Analysis Intelligence Index18.0%estimated ± 3.0 pp, high confidence
CritPt0.4%estimated ± 4.0 pp, high confidence
GPQA Diamond77.7%estimated ± 1.2 pp, high confidence
HLE16.9%estimated ± 5.1 pp, medium confidence
IFBench62.5%estimated ± 9.2 pp, medium confidence
LiveCodeBench v651.8%measured
AA-Omniscience Accuracy27.5%estimated ± 6.7 pp, low confidence
SciCode17.7%estimated ± 4.9 pp, medium confidence
IFEval88.7%measured
AA Coding Index6.7%estimated ± 5.7 pp, medium confidence
SWE-bench Verified57.5%estimated ± 9.7 pp, low confidence
SWE-bench Pro0.0%estimated ± 6.0 pp, low confidence
MMLU-Pro83.0%measured
AA-IFBench44.0%estimated ± 8.8 pp, medium confidence
AIME 202480.2%estimated ± 2.6 pp, high confidence
BBH97.7%estimated ± 2.8 pp, medium confidence
FailureSensorIQ60.8%measured
GPQA77.5%measured
MMLU89.8%estimated ± 2.1 pp, high confidence
Vals GPQA Diamond75.3%estimated ± 4.5 pp, high confidence
Vals MMLU-Pro82.1%estimated ± 2.3 pp, high confidence
MATH-50095.9%estimated ± 2.0 pp, high confidence
MMMLU86.0%estimated ± 1.7 pp, high confidence
AIME 202570.3%measured
AA MMLU-Pro85.8%estimated ± 1.2 pp, low confidence
ARC-AGI-143.5%estimated ± 6.7 pp, low confidence
ARC-AGI-20.0%estimated ± 10.9 pp, low confidence
HLE w/o tools12.6%estimated ± 5.3 pp, medium confidence
SWE Multilingual65.7%estimated ± 3.2 pp, low confidence
BinaryAudit8.2%estimated ± 8.9 pp, low confidence
OTelBench3.5%estimated ± 7.7 pp, low confidence
C-Eval89.1%estimated ± 1.0 pp, low confidence
HMMT Feb 202564.0%estimated ± 2.9 pp, medium confidence
HMMT Feb 202635.2%estimated ± 11.0 pp, low confidence
INCLUDE81.6%estimated ± 1.4 pp, low confidence
MAXIFE66.4%estimated ± 6.5 pp, low confidence
MMLU-ProX79.4%measured
MMLU-Redux92.1%estimated ± 0.6 pp, high confidence
NOVA-6354.9%estimated ± 1.3 pp, low confidence
PolyMath38.4%estimated ± 11.4 pp, low confidence
SuperGPQA62.6%measured
SimpleQA54.3%measured